Understanding precisely how ultra-processed foods damage the cardiovascular system has long been a missing piece in nutrition science. Diet questionnaires can capture what people eat, but not the molecular cascade that follows. A new multi-cohort metabolomics study closes that gap meaningfully, mapping a blood-based biological fingerprint that connects ultra-processed food consumption to coronary heart disease and death — and doing so across racially and demographically diverse populations.
The research used untargeted metabolomic profiling across three large prospective U.S. cohorts totaling nearly 7,700 adults free of cancer and cardiovascular disease at baseline. Starting from a panel of approximately 1,100 circulating metabolites, investigators identified 142 candidates associated with ultra-processed food intake in a discovery cohort (the Southern Community Cohort Study, N=1,688), then validated 43 of those metabolites in two independent cohorts — the PLCO trial (N=2,315) and the Atherosclerosis Risk in Communities Study (N=3,682). Ultra-processed food intake was classified using the NOVA framework. A composite metabolite signature derived from these validated metabolites was further associated with incident coronary heart disease, cardiovascular mortality, and all-cause mortality, with associations surviving a false discovery rate threshold below 10%.
This work is significant for several reasons beyond confirming a UPF-CVD link that observational epidemiology has repeatedly suggested. By anchoring dietary exposure to an objective biological readout, the study begins to answer the mechanistic question that has made UPF research contentious — critics have argued that observed risks might reflect confounding lifestyle factors rather than the foods themselves. A validated metabolite signature partially addresses that challenge. Importantly, the multi-cohort design with cross-validation provides stronger inference than single-cohort studies, though causality remains unestablished given the observational architecture. The specific identity of the 43 validated metabolites — whether lipid species, inflammatory markers, or food additive metabolites — will be critical for future mechanistic work. This study qualifies as a meaningful step forward, moving UPF research from epidemiological association toward biological plausibility.
